Replying to alphakamp

I did keep a copy of those dotfiles! So if you wanted to try to trouble shoot... I also tried doing gossip rebuild indexes, but it failed due to the relays table being corrupt as far as I could tell
Avatar
Mike Dilger ☑️ 2y ago

If you were getting MdbErrors then probably the database was corrupted.

I take backups of my database because I often break it as I write new code that doesn't work quite right yet:

tar cvf - lmdb | zstd -c > lmdb.$(date +%Y%m%d).zst

Reply to this note

Please Login to reply.

Discussion

cc
alphakamp 2y ago

I think so. mdb_dump -s relays ./

mdb.c:5966: Assertion 'IS_LEAF(mp)' failed in mdb_cursor_next()

Aborted (core dumped)

Avatar
Mike Dilger ☑️ 2y ago

That isn't necessarily an indication of corrupt data. It could be a bug. That comes up, for example, if you delete records from a cursor you are iterating over (which should work but doesn't and IMHO is a deep LMDB bug never corrected because it requires a specific data state that you, congratulations, stumbled upon!). But if you hard booted your machine, it might also be corrupted. I'm not keen to try to restore it to find a bug that might not be there.

cc
alphakamp 2y ago

I'm with ya, if I come across it otherwise I'll open a GitHub issue